Research map / Large language models / Agents, code and math
OmniDocLayout: Towards Diverse Document Layout Generation via Coarse-to-Fine LLM Learning
Hengrui Kang, Zhuangcheng Gu, Zhiyuan Zhao, Zichen Wen, Bin Wang, Weijia Li, Conghui He
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add coarse-grained pretraining on OmniDocLayout-1M before fine adaptation (ablation) | OmniDocLayout-LLM 0.5B · M6Doc Newspaper U-Cond | FID: 42.98 -> 39.73 | Improved |
| add coarse-grained pretraining before fine adaptation (ablation) | OmniDocLayout-LLM 0.5B · M6Doc Newspaper U-Cond | Overlap: 8.308 -> 0.084 | Improved |
| add coarse-grained pretraining before fine adaptation (ablation) | OmniDocLayout-LLM 0.5B · M6Doc Newspaper Refinement | FID: 22.07 -> 10.60 | Improved |
| add fine-grained adaptation after coarse pretraining (ablation) | OmniDocLayout-LLM 0.5B · M6Doc Newspaper U-Cond | FID: 249.1 -> 39.73 | Improved |
| scale model from 0.5B to 3B (ablation) | OmniDocLayout-LLM · M6Doc Newspaper C->S+P | FID: 10.71 -> 26.88 | Hurt |
| scale model from 0.5B to 1.5B (ablation) | OmniDocLayout-LLM · M6Doc Newspaper C->S+P | FID: 10.71 -> 10.63 | No clear effect |
| scale model from 0.5B to 3B (ablation) | OmniDocLayout-LLM · M6Doc Newspaper U-Cond | FID: 39.73 -> 35.81 | No clear effect |
| use OmniDocLayout-LLM instead of diffusion-based LayoutDM/LACE | document layout generation · M6Doc Textbook U-Cond | FID: 180.25 -> 40.28 | Improved |
| use OmniDocLayout-LLM instead of general-purpose LLMs in 0-shot | document layout generation · M6Doc Newspaper U-Cond | FID: 171.01 -> 39.73 | Improved |
| increase in-context shots for general-purpose LLMs | GPT-4o, Gemini-2.5-Flash, Claude-3.7-Sonnet · M6Doc | FID: GPT-4o textbook U-Cond 135.32 -> 71.56 | Improved |
| increase shots beyond 5 for general LLMs | general-purpose LLMs · M6Doc | cost/quality | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- LayoutPrompter — LayoutPrompter: Awaken the Design Ability of Large Language Models (NeurIPS 2023) · better on 5 of 5 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- A Simple yet Effective Layout Token in Large Language Models for Document Understanding (CVPR 2025 · similarity 0.87)
- LayoutLLM: Layout Instruction Tuning with Large Language Models for Document Understanding (CVPR 2024 · similarity 0.87)
- M2Doc: A Multi-Modal Fusion Approach for Document Layout Analysis (AAAI 2024 · similarity 0.87)
- Graph-based Document Structure Analysis (ICLR 2025 · similarity 0.87)
- AnyDoc: Enhancing Document Generation via Large-Scale HTML/CSS Data Synthesis and Height-Aware Reinforcement Optimization (CVPR 2026 · similarity 0.86)
- Relation-Rich Visual Document Generator for Visual Information Extraction (CVPR 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
